[
https://issues.apache.org/jira/browse/MYNEWT-555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15833630#comment-15833630
]
ASF subversion and git services commented on MYNEWT-555:
--------------------------------------------------------
Commit 6410e4375caf15bb265cf4a641f8e669e45bbd8c in incubator-mynewt-core's
branch refs/heads/develop from [~ccollins476]
[ https://git-wip-us.apache.org/repos/asf?p=incubator-mynewt-core.git;h=6410e43
]
MYNEWT-555 BLE Host - Avoid mbuf copy on chr read
When the host receives any form of characteristic read request, an
extraneous mbuf gets allocated. The sequence is:
1. Allocate new mbuf.
2. Pass new mbuf to application code so app can fill in
characteristic value.
3. Copy characteristic value out of new mbuf into response mbuf
4. Free new mbuf
The reason for the extra allocation is to simplify reads that specify
non-zero offsets. In the case of a GATT read, the peer may read from
different offsets of the characteristic. In an effort to simplify the
API, I decided the requested offset should be hidden from the
application. Instead, the application always provides the full
characteristic value, and the host copies the requested portion out of
the user-filled mbuf and into the ACL data packet.
The fix is to try to reuse the response to hold the characteristic data.
This can be done if the read request specifies an offset of 0. For
non-zero offsets, the old behavior remains: allocate an extra mbuf and
copy the data.
Additional context:
https://lists.apache.org/thread.html/78358745716cac9512b3c659dfe9b9b7bbb1354ee4c38d6da137bcc0@%3Cdev.mynewt.apache.org%3E
> BLE Host - Avoid mbuf copy on chr read
> --------------------------------------
>
> Key: MYNEWT-555
> URL: https://issues.apache.org/jira/browse/MYNEWT-555
> Project: Mynewt
> Issue Type: Bug
> Components: Nimble
> Reporter: Christopher Collins
> Assignee: Christopher Collins
> Fix For: v1_0_0_rel
>
>
> When the host receives any form of characteristic read request, an extraneous
> mbuf gets allocated. The sequence is:
> * Allocate new mbuf.
> * Pass new mbuf to application code so app can fill in characteristic value.
> * Copy characteristic value out of new mbuf into response mbuf
> * Free new mbuf
> The reason for the extra allocation is to simplify reads that specify
> non-zero offsets. In the case of a GATT read, the peer may read from
> different offsets of the characteristic. In an effort to simplify the API, I
> decided the requested offset should be hidden from the application. Instead,
> the application always provides the full characteristic value, and the host
> copies the requested portion out of the user-filled mbuf and into the ACL
> data packet.
> The fix is to try to reuse the response to hold the characteristic data.
> This can be done if the read request specifies an offset of 0. For non-zero
> offsets, the old behavior remains: allocate an extra mbuf and copy the data.
> Additional context:
> https://lists.apache.org/thread.html/78358745716cac9512b3c659dfe9b9b7bbb1354ee4c38d6da137bcc0@%3Cdev.mynewt.apache.org%3E
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
